The Caco-2 cell bioassay for iron (Fe) bioavailability represents a cost-effective and versatile approach to assess Fe bioavailability from foods, food products, supplements, meals, and even diet regimens. Thoroughly validated to human studies, it represents the state of the art for studies of Fe bioavailability.
Knowledge of Fe bioavailability is critical to the assessment of the nutritional quality of Fe in foods. In vivo measurement of Fe bioavailability is limited by cost, throughput, and the caveats inherent to isotopic labeling of the food Fe. Thus, there exists a critical need for an approach that is high-throughput and cost-effective. The Caco-2 cell bioassay was developed to satisfy this need. The Caco-2 cell bioassay for Fe bioavailability utilizes simulated gastric and intestinal digestion coupled with culture of a human intestinal epithelial cell line known as Caco-2. In Caco-2 cells, Fe uptake stimulates the intracellular formation of ferritin, an Fe storage protein easily measured by enzyme-linked immunosorbent assay (ELISA). Ferritin forms in proportion to Fe uptake; thus, by measuring Caco-2 cell ferritin production, one can assess intestinal Fe uptake from simulated food digests into the enterocyte.
Via this approach, the model replicates the key initial step that determines food Fe bioavailability. Since its inception in 1998, this model approach has been rigorously compared to factors known to influence human Fe bioavailability. Moreover, it has been applied in parallel studies, with three human efficacy studies evaluating Fe biofortified crops. In all cases, the bioassay correctly predicted the relative amounts of Fe bioavailability from the factors, crops, and overall diet. This paper provides detailed methods on Caco-2 cell culture coupled with the in vitro digestion process and cell ferritin ELISA necessary to conduct the Caco-2 cell bioassay for Fe bioavailability.
To fully understand the research need and benefit of the Caco-2 cell bioassay for Fe bioavailability, one must first understand the approaches that were in place prior to the advent of this model. The measurement of Fe bioavailability from a food or meal in vivo is a challenging task, particularly when combinations of food need to be assessed in a meal or diet. Isotopic labeling has been the most common approach for the measurement of Fe bioavailability over the past 50 years1. Isotopic labeling is used for single-meal and multiple-meal studies and is impractical for long-term studies. Stable isotopes of Fe such as 57Fe and 58Fe are the most commonly used; however, studies have been conducted with radioisotopes such as 59Fe, utilizing whole-body counting2. For plant foods, isotopic labeling has been done via extrinsic or intrinsic labeling. For extrinsic labeling, a known amount of isotope is added to the food or meal. The food is then mixed, and a 15-30 min equilibration period is incorporated into the protocol prior to the consumption. Hydroponic culture-adding the isotope to the nutrient solution to incorporate it into the plant while it grows and develops-is required for the intrinsic labeling of plant foods. The pros and cons of each approach are discussed below.
Extrinsic isotopic labeling
In the early to mid-1970s, human Fe absorption was studied by extrinsic labeling of Fe in foods, wherein a known amount of isotope is added to the known amount of Fe in the food or meal, mixed, and equilibrated for 15-30 min before measurements. Various amounts of extrinsic isotopes have been used, ranging from 1% to 100% of the intrinsic Fe, but most commonly in the range of 7%-30%3. Extrinsic labeling is based on the assumption that the extrinsic Fe isotope gets fully equilibrated with the intrinsic Fe of the food or meal. Extrinsic isotope absorption is then measured, and each atom of the extrinsic isotope is calculated to represent a given number of intrinsic Fe atoms. This calculation is based on the relative molar amounts. In 1983, multiple validation studies of the technique were summarized in a review paper4. Validation of the technique was done by simultaneously comparing the percent absorption of the extrinsic isotopic label to the percent absorption of an intrinsic isotopic label. Thus, a ratio of the extrinsic to intrinsic absorption close to 1 suggests that each pool of Fe was equally absorbed. At the time, a ratio close to 1 was also considered to represent equilibration of the extrinsic isotope with the intrinsic Fe of the food or meal. Ratios of extrinsic to intrinsic Fe absorption ranged from mean values of 0.40 to 1.62, with a mean (±SD) ratio of 1.08 ± 0.14 in 63 comparisons. It is important to note that, in all of the studies summarized in this review, none directly tested the equilibration of the extrinsic label with the intrinsic Fe. In summary, the authors of the review concluded the following:
"The extrinsic tag technique has proven valid for several foods under certain experimental conditions. But, this method cannot yet be considered proven with regards to all types of foods. The extrinsic tag method is not appropriate for monitoring iron absorption from a diet that contains insoluble forms of iron. The validity of this technique relies upon the basic assumption that the extrinsic tag exchanges completely with all endogenous nonheme food iron. At present it is not known how completely the different forms of nonheme iron are labeled by an extrinsic tag. This is important in light of studies which have suggested that iron inhibitors may affect the extrinsic tag differently than some forms of nonheme iron in foods. Research on food factors which can impair a complete isotopic exchange is scant. Thus, interpretation of bioavailability data from extrinsic tag research requires consideration of inhibitors of exchange which may be present in the food or diet."
Since 1983, only two studies have been published that evaluated the accuracy of extrinsic labeling of Fe3,5. In both these studies, the equilibration of an extrinsic isotopic label was directly compared with the intrinsic Fe of the foods, which, in these studies, were staple food crops. White, red, and black bean varieties were tested, along with lentils and maize. Using established in vitro digestion techniques and the measurement of Fe solubility and precipitation, both studies demonstrated that extrinsic isotopic labeling does not consistently result in full equilibration, with evidence that, for some bean varieties, the misequilibration can be very high depending on the amount of extrinsic isotope and seed coat color3. Despite the conclusions of the 1983 review paper, extrinsic labeling studies of beans continued6,7,8,9,10,11,12. None of these studies included testing the equilibration of the extrinsic label with the intrinsic Fe.
Intrinsic labeling
Intrinsic labeling of plant food for the assessment of Fe bioavailability eliminates the accuracy issues of equilibration in extrinsic labeling. However, this approach cannot yield large amounts of material because of the requirement of greenhouse space for hydroponic culture. Hydroponic culture is labor-intensive, requires a high quantity of expensive stable isotope, and often results in plant growth different in terms of yield and seed Fe concentration. Due to the cost, intrinsic labeling is only suitable for small-scale studies aimed at understanding mechanisms underlying Fe uptake or factors influencing Fe uptake from foods. Production of 1-2 kg of a staple food crop costs approximately $20,000-$30,000 for materials alone, depending on the isotope and hydroponic approach13,14.
Given the challenges associated with isotopic labeling, investigators sought to develop in vitro approaches. Early methods utilized simulated gastric and intestinal food, coupled with the measurement of Fe solubility or Fe dialyzability as an estimate of bioavailability15. Such studies quickly found that Fe dialyzability was not a consistent measure of bioavailability as Fe can be soluble, tightly bound to compounds and, therefore, not exchangeable, leading to the overestimation of bioavailability. To address these issues, methodology to utilize a human intestinal cell line evolved, thereby adding a living component and enabling the measurement of Fe uptake16. The human intestinal cells-Caco-2 cells-originated from a human colon carcinoma and have been widely used in nutrient uptake studies. This cell line is useful as, in culture, the cells differentiate into enterocytes that function similarly to the brush border cells of the small intestine. Studies have shown that Caco-2 cells exhibit the appropriate transporters and response to factors that influence Fe uptake17,18.
The initial studies, utilizing radioisotopes to measure Fe uptake in Caco-2 cells, were refined to measure Fe uptake based on Caco-2 cell ferritin formation. Caco-2 cell ferritin measurement enhanced sample throughput and negated issues of radioisotope handling and the equilibration of extrinsic Fe with intrinsic Fe19,20. Measurement of Fe uptake via ferritin formation enabled researchers to study a broad range of foods, including complex meals21. Thus, simulated (in vitro) digestion coupled with Caco-2 cell Fe uptake provided a better physiological assessment of Fe uptake from foods. It is important to note that this model primarily determines relative differences in Fe bioavailability. Like many useful cell lines, Caco-2 cells also have shown variability in responsiveness but have maintained consistent relative differences in Fe uptake between foods. Proper technique and careful attention to detail can improve consistent cell ferritin formation response in Caco-2 cells.
The in vitro digestion/Caco-2 cell model is also known as Caco-2 cell bioassay. This assay has been thoroughly validated via direct comparison to human and animal studies22. In addition to the direct parallel comparison of the bioassay to human efficacy trials, this model has been shown to exhibit a qualitatively similar response in Fe uptake to that of humans18,19,23. Therefore, as an in vitro approach, the Caco-2 cell bioassay warrants high credibility as a screening tool for evaluating Fe nutrition from foods. It has been widely applied to numerous foods and food products21,24,25,26,27,28.
Since its inception in 1998, the Caco-2 cell bioassay has advanced the field of Fe nutrition as it has helped identify factors that influence intestinal Fe uptake. In so doing, this model has developed and refined research objectives for more definitive and less costly human studies. One could also argue that the use of the model negates the need for some human trials.
In summary, the relative delivery of Fe from a food or meal can be measured with the Caco-2 cell bioassay. Regardless of the amount of Fe in the test meal, the bioassay defines the relative amount of Fe taken up into the enterocyte-the first step of the absorption process. This is the most important step in defining Fe bioavailability, as most often the goal is to measure with the intent to improve or, at the very least, monitor the nutritional quality of Fe in a food. Given that iron status is regulated by absorption, and thus Fe uptake is upregulated in Fe-deficient individuals to meet nutritional needs, the standard conditions of the model are designed so that Fe uptake by the cells is maximal. In this way, the bioassay provides a true measure of the potential of the food to deliver Fe.
NOTE: As a convenient point of reference for readers, the following methodology describes the specific culture conditions and materials required for the measurement of Fe bioavailability from 20 experimental samples, plus the required quality controls, in a run of the bioassay. Increasing the number of samples beyond this capacity is not recommended due to the time required for various cell culture and in vitro digestion steps within the bioassay.
1. Choosing the amount of samples
2. Preparation of samples
3. Caco-2 cell culture
4. In vitro digestion
5. Measurement of Caco-2 cell ferritin and cell protein
Identification and measurement of Fe bioavailability in staple food crops
One of the primary reasons for developing this model was to identify factors that influence Fe bioavailability in staple food crops and provide a tool for plant breeders that would enable them to identify and develop varieties with enhanced Fe bioavailability. The common bean (Phaseolus vulgaris) has been targeted globally as a crop for Fe biofortification; thus, the model has been applied extensively to evaluate the nutritional quality of Fe in a broad range of bean market classes and bean breeding programs. For example, yellow beans are an emerging market class in the United States. In regions such as East Africa, they are highly popular and are widely known to be fast-cooking and considered by many to be "easy to digest." Recent studies with the Caco-2 cell bioassay have demonstrated that certain varieties of yellow beans can have high Fe bioavailability relative to other color classes (Figure 2). In this study, the Manteca varieties were identified as being high in Fe bioavailability relative to reference controls of the white and red-mottled color classes. Moreover, the results were consistent across two consecutive harvest years. Such comparisons are simply not feasible in other models, particularly in vivo models, due to the high cost and much lower throughput of animal and human trials.
Evaluation of food processing effects on Fe bioavailability
The Caco-2 cell bioassay can also be applied to evaluate food processing effects on Fe bioavailability. For example, the results in Figure 3 are from an analysis of beans and bean-based pasta of multiple color classes. The results demonstrate how processing the beans into a flour increased Fe bioavailability from white (Snowdon, Alpena, and Samurai) and yellow (Canario) bean varieties. For the cranberry (Etna), red kidney (Red Hawk), and black (Zenith) varieties, Fe bioavailability decreased in the pasta flour preparations. Related analyses demonstrated that processing the beans into a flour disrupted the cotyledon cell walls of the beans, thus making the intracellular Fe accessible for uptake. Iron uptake increased in the white and yellow bean pasta as the seed coats of these varieties did not contain polyphenolic compounds that inhibit Fe bioavailability. In contrast, the seed coats of cranberry, red kidney, and black beans contain high levels of inhibitory polyphenols, thus decreasing Fe uptake. These results clearly indicate the usefulness of the model in exposing factors that can influence the nutritional quality of Fe, which otherwise would go undetected.
Figure 1: Insert ring setup for Caco-2 cell Fe uptake. (A) Image of Caco-2 cells and insert ring with attached dialysis membrane. (B) Diagram of the overall procedure for in vitro digestion coupled with Caco-2 cell Fe uptake within a single well of the multi-well plate. Please click here to view a larger version of this figure.
Figure 2: Iron bioavailability scores of unsoaked and cooked whole-seed genotypes in a diverse panel of yellow beans. (A) Field season 2015; (B) field season 2016. Values are means (standard deviation) of triplicate measurements from two field replicates per genotype (n = 6). Genotypes are categorized on the x-axis by cooking class, ranked from the fastest-cooking genotype to the slowest-cooking entry. *Significantly lower (p < 0.05) iron bioavailability score than the other YBP entries. **Significantly higher (p < 0.05) iron bioavailability scores than the other YBP genotypes. This figure was modified from32. Please click here to view a larger version of this figure.
Figure 3: Iron bioavailability expressed as Caco-2 cell ferritin formation (nanogram of ferritin per milligram of cell protein) of bean varieties and their corresponding bean-based spaghettis. (A) Three white bean varieties and their corresponding bean-based spaghettis; (B) four colored bean varieties and their corresponding bean-based spaghettis. Values are the means (± standard deviation) of six measurements from each variety. The blue hyphenated line indicates the iron bioavailability of a non-fortified durum wheat pasta control extruded, cooked, and processed in the same manner as the bean-based spaghettis. *Significantly (p ≤ 0.05) higher Caco-2 cell ferritin formation than whole beans after cooking. **Significantly (p ≤ 0.05) lower Caco-2 cell ferritin formation than whole beans after cooking. This figure was modified from33. Please click here to view a larger version of this figure.
Since its inception, numerous studies have been published that describe this method for the Caco-2 cell bioassay. The basic conditions have remained relatively unchanged since the initial publication in 199818. However, over the past 20 years, numerous technical details have been refined and standardized to yield unprecedented consistency in the response of the bioassay. Careful and precise adherence to the cell culture and in vitro digestion conditions are the key to the consistent and sensitive response of the bioassay.
From our experience in training numerous individuals in the use of this method, the most common struggle is the proper culture of the Caco-2 cells. Consistent culture of healthy monolayers is key to healthy and responsive Caco-2 cell monolayers. If cell protein levels are not highly consistent from well to well and not within the range of cell protein listed in the protocol, the investigator should reexamine the cell culture conditions for deviation from the protocol. Alternatively, low-level microorganism contamination may exist, the cell culture incubator may not be operating properly, or the cell culture medium may not be properly formulated.
The in vitro digestion process is another source of potential problems. Removal of contaminant Fe from the digestive enzymes is critical. Despite manufacturer claims, it is prudent to periodically check the Fe concentration of the enzymes and make sure that the Fe removal process (see protocol) is effective. If Fe contamination is present in the digestive enzymes, then the baseline digest quality control will yield cell ferritin values in excess of the recommended range.
An experienced and trained investigator should be able to analyze 20 experimental samples, plus the quality controls, in a single run of the bioassay. Thus, approximately 12 six-well plates are needed for each run of the bioassay. Higher numbers of samples per bioassay are not recommended as the timing for pH titration during the digestion process can be too long, leading to potential inconsistency between sample digestion times.
The disadvantages of this model are relatively few. It requires investigators who are highly skilled in cell culture and capable of precise attention to detail and protocol. Laboratory space must be clean of sources of Fe contamination, and reagents and other materials should be routinely monitored for Fe contamination. Thus, the user should have the capability or access to instrumentation for the measurement of Fe concentration. This model is only a relative or semi-quantitative measure. However, with the proper use of reference controls, the model can provide some quantitative estimates of Fe absorption. Indeed, a conversion equation of absorption ratios of control versus test material has been generated19.
The bioassay works according to the following principle: Caco-2 cells produce more ferritin protein in response to increases in cellular Fe concentrations. Therefore, Fe bioavailability is proportional to the increase in Caco-2 cell ferritin production. This increase is expressed as a ratio of cell ferritin to total Caco-2 cell protein (nanogram of ferritin per milligram of total cell protein) after exposure to a digested sample19. Ferritin measurements are made using an ELISA kit (see the Table of Materials) tested for response in this bioassay. Total cell protein concentrations are quantified using a protein assay kit. As mentioned previously, under the conditions used for this method, typical Caco-2 cell protein levels in a six-well plate range from 2.0 mg to 2.6 mg of cell protein per well. Values outside of this range indicate unhealthy cell cultures, possible overgrowth of cells, or poor cell seeding technique. Within a given run of the bioassay, values should only vary up to 0.2 mg per well. Furthermore, under the seeding densities and culture conditions used in this methodology, there is substantial brush border enzyme activity at 13 days post seeding, indicating maturation of most, if not all, of the cell monolayer28,29,30. Monitor cell monolayers throughout the 13 days prior to use for contamination or stress, such as vacuole formation or gaps in monolayer formation. If such conditions are evident, the cells should not be considered valid for use in the bioassay.
To monitor the responsiveness of the Caco-2 bioassay, each experiment should be run with several quality controls, including a blank digest, containing only the physiologically balanced saline and the gastrointestinal enzymes. These controls ensure that there is no Fe contamination in the bioassay. Ferritin values of Caco-2 cells exposed to the blank digest typically range from 1 ng to 6 ng of ferritin/mg cell protein. Baseline ferritin in this range also indicates that the cells are at relatively low Fe status and, thus, should exhibit maximal sensitivity to available Fe.
For the initial 15 years of use, additional quality controls included 1) a blank digest with FeCl3 (66 µM) and 2) a blank digest of FeCl3 (66 µM) plus the addition of 1.3 mM ascorbic acid. Ferritin values for the FeCl3 digest were typically in the range of 30-50 ng of ferritin/mg cell protein, and the FeCl3 digest with ascorbic acid was in the range of 250-400 ng of ferritin/mg cell protein. In more recent years, the blank digest remains a quality control; however, we have switched to using a food sample with and without ascorbic acid at a ratio of 20:1, ascorbate:Fe. The food sample used was a white bean flour that contains approximately 65 µg Fe/g of sample. These quality controls give a narrower and more consistent range of response, yielding 20-30 ng of ferritin/mg of cell protein for the white bean flour and 70-150 ng of ferritin/mg of cell protein for the white bean flour plus ascorbate. It should be noted that the new range of values are from the referenced kit in the Table of Materials, which tends to be slightly lower than the now-defunct Ramco ELISA kit. As of the publication of this manuscript, only 2-3 months of data have been acquired with the referenced kit.
It is important to recognize the results and cell culture conditions that indicate an invalid or suboptimal run of the bioassay. First, as stated in the methods, if the blank digest conditions yield cell ferritin concentrations higher than the suggested range, this could be indicative of Fe contamination of the cell culture media, the digestive enzymes, or the dialysis membrane. Acceptable Fe concentrations for the cell culture media and the digestive enzymes are <20 µg Fe/mL. Values outside the range for the other quality controls, particularly if they are on the low side, also indicate that the validity of the results is questionable.
In summary, this model is highly sensitive to bioavailable Fe, as the cell culture conditions are designed to create cells of low Fe status; thus, their mechanisms for Fe uptake are highly upregulated. It is a robust model capable of high throughput. Any food or diet that can be fed to humans can be assessed in this model and, therefore, the bioassay has a broad range of applications. Plant breeders can use this model to measure Fe bioavailability in staple foods, identifying traits and chromosomal regions that affect Fe bioavailability. Food scientists can apply the model to determine optimal formulations and evaluate the effects of processing to ensure adequate Fe bioavailability. Nutritionists can use the model to evaluate and monitor dietary Fe bioavailability from individual foods, food combinations, and even diet plans. It has been thoroughly validated to human trials, correctly predicting the direction and magnitude of effects in every application. Thus, by combining simulated digestion with intestinal epithelial cell Fe uptake, this model represents the critical first step in the Fe absorption process and is, therefore, capable of predicting the delivery or bioavailability of Fe from foods.
The authors have nothing to disclose.
The author is deeply grateful for the technical efforts of Yongpei Chang and Mary Bodis. The extremely successful application of this model in the field of nutrition is a direct result of their expertise and attention to detail. The development of this model was funded entirely by the United States Department of Agriculture, Agricultural Research Service.
0.5 M HCl | Fisher Scientific | A508-4 Hydrochloric Acid TraceMetal Grade | |
18 megaohm water | Also known as distilled, deionized water | ||
3,3′,5-Triiodo-L-thyronine sodium salt | Sigma Aldrich Co | T6397 | |
6-well plates | Costar | 3506 | Use for bioassay experiments |
ascorbic acid | Sigma Aldrich Co | A0278 | |
bile extract | Sigma Aldrich Co | B8631 | |
Caco-2 cells | American Type Culture Collection | HTB-37 | HTB-37 is a common variety. |
Cell culture flasks T225 | Falcon | 353138 | |
Cell culture flasks T25 | Corning | 430639 | |
Cell culture flasks T75 | Corning | 430641U | |
Chelex-100 | Bio-Rad Laboratories Inc | 142832 | Known as the weak cation exchange resin in the protocol |
collagen | Corning | 354236 | |
dialysis membrane | Spectrum Laboratories | Spectra/Por 7 Pretreated RC Dialysis Tubing 15,000 MWCO | Spectra/Por 7 Pretreated RC Dialysis Tubing 15,000 MWCO |
Dulbecco’s Modified Eagle’s Medium | Gibco | 12100046 | DMEM |
epidermal growth factor | Sigma Aldrich Co | E4127-5X.1MG | |
Ferritin ELISA Assay Kit | Eagle Biosciences | FRR31-K01 | |
fetal bovine serum | R&D Systems | S12450 | Optima |
HEPES | Sigma Aldrich Co | H3375 | |
Hydrocortisone-Water Soluble | Sigma Aldrich Co | H0396 | |
insert ring | Corning Costar | not sold | Transwell, for 6 well plate, without membrane |
insulin | Sigma Aldrich Co | I2643 | |
KCl | Sigma Aldrich Co | P9333 | |
large column | VWR International | KT420400-1530 | |
Minimum Essential Medium | Gibco | 41500034 | MEM |
NaCl | Fisher Scientific | S271 | |
pancreatin | Sigma Aldrich Co | P1750 | |
PIPES disodium salt | Sigma Aldrich Co | Piperazine-1,4-bis(2-ethanesulfonic acid) disodium salt P3768 | |
porcine pepsin | Sigma Aldrich Co | P6887 or (P7012-25G Sigma | |
protein assay kit | Bio-Rad Laboratories Inc | Bio-Rad DC protein assay kit 500-0116 | Measurement of Caco-2 cell protein |
silicone o rings | Web Seal, Inc Rochester NY | 2-215S500 | |
sodium bicarbonate | Fisher Scientific | S233 | |
Sodium selenite | Sigma Aldrich Co | S5261 | |
ZellShield | Minerva Biolabs | 13-0050 | Use at 1% as antibiotic/antimycotic ordered through Thomas Scientific |